3
\$\begingroup\$

In circuit designing, it is a common phenomenon to presume that in case of nMOS the channel current flows from drain to source (also seen in schematics), while in the case of pMOS, channel current flows from source to drain.

What characteristic in MOSFETs coerces this distinction? Is it simply something to do with fabrication?

Because, as far as I understand, the distinction between the two comes from the substrate/wells (p-type/n-type in case of nMOS and n-type/p-type in the case of pMOS) and the gate voltages (+ve for nMOS and -ve for pMOS), but once the channel is formed then the channels along with the wells and the source drain contacts should be able to basically behave like a resistor, allowing channel current flow in both directions.

\$\endgroup\$
6

6 Answers 6

10
\$\begingroup\$

When talking about MOSFETs on an IC, these mostly have a symmetrical structure like this:

enter image description here

Picture from here.

Note how Drain and Source are physically identical.

When using these type of MOSFETs (so when designing a circuit that will be etched on a chip!) then the direction of the current defines which area we call source and which we call drain.

In some situations that direction is unclear (typical example is a "passgate circuit" which is a bidirectional switch), then it is up to the designer to choose which side is the source and which is the drain.

The explanation above does not apply to discrete MOSFETs (and also on-chip High voltage MOSFETs) as these have a non-symmetrical structure. Here's an example of such a Vertical MOSFET:

enter image description here

In such a MOSFET the Drain is constructed such that is can withstand a high voltage and/or it can dissipate a lot of power by mounting the device with the drain side directly on the metal case of the MOSFET and mounting the MOSFET on a heatsink.

\$\endgroup\$
5
\$\begingroup\$

They can conduct in both directions, but you have to take into account also the intrinsic diode.

Example of N-MOSFET. They are both correct, except that 2nd is almost useless since it won't work as ON/OFF switch.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
1
  • 4
    \$\begingroup\$ I think we should note that this applies only to discrete three-terminal devices, where the body has been intentionally connected to one terminal, the source. \$\endgroup\$ May 30, 2021 at 13:23
3
\$\begingroup\$

it is a common phenomenon to presume...

To presume something wouldn't be regarded as a phenomenon (not in humans anyway).

Channels can carry current in either direction because there is nothing in particular about the channel of a MOSFET that makes it uni-directional to current. MOSFETs generally have a 4-quadrant response to the applied drain-source voltage: -

enter image description here

Picture from here.

\$\endgroup\$
1
  • 2
    \$\begingroup\$ I think we should note that this applies only to discrete three-terminal devices, where the body has been intentionally connected to one terminal, the source. \$\endgroup\$ May 30, 2021 at 13:34
3
\$\begingroup\$

I think the part of the discussion you are missing is that for a generic, four-terminal MOSFET it is possible for the source and drain to be swapped depending on the applied voltage. For an NMOS transistor, the source is by definition the terminal at the lower voltage so current always flows from drain to source. For a PMOS transistor, the source is always by definition the terminal at the higher voltage so current always flow from source to drain. If you think about how a bidirectional transmission gate works in CMOS VLSI design you can see this behavior, as the notion of "source" and "drain" flips when the direction of current flow reverses.

Sometimes MOSFETs are packaged as discrete devices with only three terminals. The MOSFET body will be connected to one of terminals, which effectively forces that terminal to be the source.

\$\endgroup\$
2
\$\begingroup\$

For a four-terminal symmetric MOSFET, the terminology of "source" and "drain" pertains to the locations where the channel carriers travel from, and to, respectively. An nFET has electrons as the majority carriers in its channel after it inverts (from weakly p-doped), and as they flow from source to drain, the conventional (positive) current is drain to source.

On the other hand, on a pFET, holes flow from source to drain in an inverted channel formed in an n-substrate or n-well, and hence the conventional current flows in the same direction.

It's worth noting the following points:

  • A four-terminal transistor may be fabricated with one of its implants doped lightly. This is called a "lightly-doped drain" technique, and provides performance/reliability benefits when that implant is used as the drain terminal. It will still operate as a MOSFET with current flow in the other direction, but will not realize the same level of performance (e.g. it may be more susceptible to hot carrier injection effects)
  • Even if a MOSFET is symmetrical, for purposes of netlisting, schematic capture, and layout, one implant is called source and one is called drain. A correct device model will, however, have correct behavior if the current flow is opposite the expected direction based on pin names.
\$\endgroup\$
0
\$\begingroup\$

The FET works as a (bad) diode in the other direction, so when you design a circuit, that defines what kind of FET you can place.

\$\endgroup\$
5
  • 3
    \$\begingroup\$ To be pedantic that’s only the case for a 3-terminal FET (which is all of them these days), but if you’re designing an IC then you have the option not to connect the substrate to the source and can then operate a FET more or less as symmetrical. \$\endgroup\$
    – Frog
    May 30, 2021 at 10:07
  • \$\begingroup\$ Incorrect, the mosfet channel is still active and significantly shunts the diode. Only when the mosfet is inactive does the diode play its role. \$\endgroup\$
    – Andy aka
    May 30, 2021 at 10:25
  • \$\begingroup\$ Yes, so you wouldn't place it in a way where an inactive FET doesn't pinch off the current. \$\endgroup\$ May 30, 2021 at 10:31
  • \$\begingroup\$ I think we should note that this applies only to discrete three-terminal devices, where the body has been intentionally connected to one terminal, the source. This does not apply to MOSFETs in general. \$\endgroup\$ May 30, 2021 at 13:24
  • \$\begingroup\$ @Frog Yes, your point is correct and important here. \$\endgroup\$ May 30, 2021 at 13:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.